Common input for all CGO solvers - Revision history

Bjorn: /* Prob structure */

2014-07-14T09:47:30Z

Prob structure

← Older revision		Revision as of 04:47, 14 July 2014
Line 67:		Line 67:
	\|''GO''\|\|Structure with parameters specific to global optimization. See the [[#GO structure\|table below]] describing the fields used in the GO structure.		\|''GO''\|\|Structure with parameters specific to global optimization. See the [[#GO structure\|table below]] describing the fields used in the GO structure.
	\|-valign="top"		\|-valign="top"
	\|''MIP''\|\|Structure with parameters specific mixed integer optimization. See the [[#MIP structure\|table below]] describing the fields used in the MIP structure.		\|''MIP''\|\|Structure with parameters specific to mixed integer optimization. See the [[#MIP structure\|table below]] describing the fields used in the MIP structure.
	\|}		\|}

Bjorn at 14:12, 20 June 2014

2014-06-20T14:12:14Z

← Older revision		Revision as of 09:12, 20 June 2014
Line 1:		Line 1:
			{{Part Of Manual\|title=the CGO Manual\|link=[[CGO\|CGO Manual]]}}

	=Input parameters=		=Input parameters=

Line 222:		Line 224:

	\|}		\|}

			[[Category:CGO]]

Bjorn: /* CGO structure */

2014-06-20T13:42:27Z

CGO structure

← Older revision		Revision as of 08:42, 20 June 2014
Line 164:		Line 164:

	\|-valign="top"		\|-valign="top"
	\|''epsDist''\|\|If ''epsDist'' > 10<sup>-7</sup>, all points x in the sphere <nowiki>\|\|</nowiki>''x'' - ''XMIN''(:,i)<nowiki>\|\|</nowiki> <= ''epsDist'' are excluded from the search.<br>Default is ''epsDist'' = 0, then the solver does not try to detect local minima, and just continues until ''maxFunc'' reached		\|''epsDist''\|\|If ''epsDist'' > 10<sup>-7</sup>, all points x in the sphere <nowiki>\|\|</nowiki>x - XMIN(:,i)<nowiki>\|\|</nowiki> <= epsDist are excluded from the search.<br>Default is ''epsDist'' = 0, then the solver does not try to detect local minima, and just continues until ''maxFunc'' reached

	\|-valign="top"		\|-valign="top"

Bjorn: /* CGO structure */

2014-06-20T13:41:50Z

CGO structure

← Older revision		Revision as of 08:41, 20 June 2014
Line 164:		Line 164:

	\|-valign="top"		\|-valign="top"
	\|''epsDist''\|\|If ''epsDist'' > 10<sup>-7</sup>, all points x in the sphere <~~math~~>\\|x - XMIN(:,i)\\| <= epsDist~~</math>~~ are excluded from the search.<br>Default is ''epsDist'' = 0, then the solver does not try to detect local minima, and just continues until ''maxFunc'' reached		\|''epsDist''\|\|If ''epsDist'' > 10<sup>-7</sup>, all points x in the sphere <nowiki>\|\|</nowiki>''x'' - ''XMIN''(:,i)<nowiki>\|\|</nowiki> <= ''epsDist'' are excluded from the search.<br>Default is ''epsDist'' = 0, then the solver does not try to detect local minima, and just continues until ''maxFunc'' reached

	\|-valign="top"		\|-valign="top"

Bjorn: /* CGO structure */

2014-06-19T18:54:13Z

CGO structure

← Older revision		Revision as of 13:54, 19 June 2014
Line 106:		Line 106:
	\|'''Field'''\|\|'''Description'''		\|'''Field'''\|\|'''Description'''
	\|-valign="center"		\|-valign="center"
	\|''Percent''\|\|rowspan="10"\| See [[CGO_expDesign\|expDesign]] for detailed information on how to set the experimental design parameters.		\|''Percent''\|\|rowspan="9"\| See [[CGO_expDesign\|expDesign]] for detailed information on how to set the experimental design parameters.
	\|-valign="top"		\|-valign="top"
	\|''nSample''		\|''nSample''
Line 124:		Line 124:
	\|''CLHMethod''		\|''CLHMethod''
	\|-valign="top"		\|-valign="top"
	\|''SCALE''		\|''SCALE''\|\|0 - Original search space (Default if any integer values)
			1 - Transform search space to unit cube (Default if no integers)
	\|-valign="top"		\|-valign="top"
	\|''objType''\|\|Tranformation of function F, all transformations are computed and saved. Selected objType is used in CGO interpolations.		\|''objType''\|\|Tranformation of function F, all transformations are computed and saved. Selected objType is used in CGO interpolations.

Bjorn at 18:48, 19 June 2014

2014-06-19T18:48:45Z

@@ Line 62: / Line 62: @@
 |-valign="top"
 |''CGO''||Structure with parameters specific to costly global optimization. See the [[#CGO structure|table below]] describing the fields used in the CGO structure.
 |-valign="top"
 |''MIP''||Structure with parameters specific mixed integer optimization. See the [[#MIP structure|table below]] describing the fields used in the MIP structure.
@@ Line 172: / Line 174: @@
 |-valign="top"
 |''localSolver''||Local optimization solver used for subproblem optimization. If not defined, the TOMLAB  default constrained NLP solver is used.
 |}

Bjorn: /* CGO structure */

2014-06-19T18:29:11Z

CGO structure

@@ Line 161: / Line 161: @@
 |-valign="top"
-|''epsDist''||If ''epsDist'' > 10<sup>-7</sup>, all points x in sphere <math>\|x - XMIN(:,i)\| <= epsDist</math> are excluded from the search.<br>Default ''epsDist'' = 0, then the solver does not try to detect local minima, and just continues until ''maxFunc'' reached
+|''epsDist''||If ''epsDist'' > 10<sup>-7</sup>, all points x in the sphere <math>\|x - XMIN(:,i)\| <= epsDist</math> are excluded from the search.<br>Default is ''epsDist'' = 0, then the solver does not try to detect local minima, and just continues until ''maxFunc'' reached
 |-valign="top"
-|''globalSolver''||Global optimization  solver used for subproblem optimization.     Default ''glcCluster  ''(SMOOTH=1)    or  ''glcDirect   ''(SMOOTH=0).       If   the  global- Solver is ''glcCluster'', the fields ''Prob.GO.maxFunc1'', ''Prob.GO.maxFunc2'', ''Prob.GO.maxFunc3'', ''Prob.GO.localSolver'', ''Prob.GO.DIRECT ''and other fields set in ''Prob.GO ''are used. See the help for these parameters in ''glcCluster''.
+|''SMOOTH''||= 1 The problem is smooth enough for local search using numerical gradient estimation methods
-'''- Special RBF  algorithm  parameters in Prob.CGO -'''
+= 0 the problem is nonsmooth or noisy, and local search methods using numerical gradient estimation are likely to produce garbage search directions.
 |-valign="top"
-|''N''||Cycle length in idea 1 (default N=1 for fStarRule 3, otherwise default N=4) or idea 2 (always N=3).
+|''globalSolver''||Global optimization  solver used for subproblem optimization.     Default [[glcCluster|''glcCluster'']]  (SMOOTH=1)    or  [[glcDirect|''glcDirect'']] (SMOOTH=0).       If   the  global- Solver is [glcCluster|''glcCluster'']], the fields ''Prob.GO.maxFunc1'', ''Prob.GO.maxFunc2'', ''Prob.GO.maxFunc3'', ''Prob.GO.localSolver'', ''Prob.GO.DIRECT ''and other fields set in ''Prob.GO ''are used. See the help for these parameters in [[glcCluster|''glcCluster'']].
 |-valign="top"
-|''infStep''||If =1, add search step with target value ''-8 ''first in cycle. Default 0. Always
+|''localSolver''||Local optimization solver used for subproblem optimization. If not defined, the TOMLAB  default constrained NLP solver is used.
 |}

Bjorn at 16:03, 19 June 2014

2014-06-19T16:03:41Z

← Older revision		Revision as of 11:03, 19 June 2014
Line 104:		Line 104:
	\|'''Field'''\|\|'''Description'''		\|'''Field'''\|\|'''Description'''
	\|-valign="center"		\|-valign="center"
	\|''Percent''\|\|rowspan="9"\| See [[CGO_expDesign\|expDesign]] for detailed information on how to set the experimental design parameters.		\|''Percent''\|\|rowspan="10"\| See [[CGO_expDesign\|expDesign]] for detailed information on how to set the experimental design parameters.
	\|-valign="top"		\|-valign="top"
	\|''nSample''		\|''nSample''
Line 122:		Line 122:
	\|''CLHMethod''		\|''CLHMethod''
	\|-valign="top"		\|-valign="top"
	\|''SCALE''\|\|~~0 - Original search space (default if any integer values)~~.		\|''SCALE''
			\|-valign="top"
			\|''objType''\|\|Tranformation of function F, all transformations are computed and saved. Selected objType is used in CGO interpolations.

	1 - ~~Transform search space to unit cube~~ (~~default~~ if ~~no integers~~).		{\|class="wikitable"
			!Value\|\|Transformation
			\|-
			\|0\|\|No transformation (default).
			\|-
			\|1\|\|F<sub>t</sub>(1) = log<sub>10</sub>(F + F<sub>add</sub>), F<sub>add</sub>=10<sup>k</sup>, where k=1,2,...lowest k to assure F + F<sub>add</sub> > 0
			\|-
			\|2\|\|F<sub>t</sub>(2) = log(1+F) if F >= 0, -log(1-F) if F < 0
			\|-
			\|3\|\|F<sub>t</sub>(3) = log(F), if all(F) > 0, otherwise NaN
			\|-
			\|4\|\|F<sub>t</sub>(4) = -log(-F) transformation, if all(F) < 0, otherwise NaN
			\|-
			\|5\|\|F<sub>t</sub>(5) = -1/y transformation, if either all(F) < 0 or all(F) > 0, otherwise NaN
			\|-
			\|6\|\|F<sub>t</sub>(6) = F(F >= median(F)) = median(F)
			\|-
			\|7\|\|F<sub>t</sub>(7) = F(F > median(F)) = median(F) + log<sub>10</sub>(1+F+median(F))
			\|-
			\|8\|\|F<sub>t</sub>(8) = F(F > 10<sup>''REPLACE''</sup>) = 10<sup>4</sup> + log<sub>10</sub>(1+F-10<sup>''REPLACE''</sup>)
			\|-
			\|-1\|\|Only for [[CGO ego\|ego solver]]. Do cross validation checking all ''objType'' on the initial experimental data and selecting the best possible.
			\|}
	\|-valign="top"		\|-valign="top"
	\|''REPLACE''\|\|0 - ~~No replacement~~, ~~default for constrained problems~~.		\|''REPLACE''\|\|In ''objType'' == 8, large function values Z are replaced by new values<br>
			Replacement: Z:= FMAX + log<sub>10</sub>(Z-FMAX+1), where FMAX = 10<sup>''REPLACE''</sup>,<br>
			Default ''REPLACE'' = 4.

	1 - ~~Large function values are replaced by the median~~.		\|-valign="top"
			\|''XMIN''\|\|Matrix d x nLocal with nLocal minima previously found. Best point(s) found might be global minima<br>Sampling of points will be avoided around these points, see epsDist below

	''> ''1 - Large values Z are replaced by new values. The replacement is defined as ''Z '':= ''FMAX ''+ ''log''10(''Z - FMAX ''+ 1), where ''FMAX ''= 10<sup>''REPLACE ''</sup>, if ''min''(''F '') ''< ''0 and ''FMAX ''= 10<sup>(''ceil''(''log''10(''min''(''F '')))+''REPLACE'')</sup>, if ''min''(''F '') ''= ''0. A new replacement is computed in every iteration, because ''min''(''F '') may change. Default REPLACE = 5, if no linear or nonlinear constraints.
	\|-valign="top"		\|-valign="top"
	\|''~~LOCAL~~''\|\|~~0 - No local searches after global search. If RBF surface is inaccurate, might be an advantage.~~		\|''FMIN''\|\|Column vector with nLocal costly function values f(x) computed at the nLocal points in ''XMIN''

	~~1 - Local search from best points after global search. If equal best function values, up to 20 local searches are done.~~
	\|-valign="top"		\|-valign="top"
	\|''~~SMOOTH~~''\|\|1 - ~~The problem is smooth enough for local search using numerical gradient estimation methods~~ (~~default~~).		\|''epsDist''\|\|If ''epsDist'' > 10<sup>-7</sup>, all points x in sphere <math>\\|x - XMIN(:,i)\\| <= epsDist</math> are excluded from the search.<br>Default ''epsDist'' = 0, then the solver does not try to detect local minima, and just continues until ''maxFunc'' reached

	~~0 - The problem is nonsmooth or noisy, and local search methods using numer- ical gradient estimation are likely to produce garbage search directions.~~
	\|-valign="top"		\|-valign="top"
	\|''globalSolver''\|\|Global optimization solver used for subproblem optimization. Default ''glcCluster ''(SMOOTH=1) or ''glcDirect ''(SMOOTH=0). If the global- Solver is ''glcCluster'', the fields ''Prob.GO.maxFunc1'', ''Prob.GO.maxFunc2'', ''Prob.GO.maxFunc3'', ''Prob.GO.localSolver'', ''Prob.GO.DIRECT ''and other fields set in ''Prob.GO ''are used. See the help for these parameters in ''glcCluster''.		\|''globalSolver''\|\|Global optimization solver used for subproblem optimization. Default ''glcCluster ''(SMOOTH=1) or ''glcDirect ''(SMOOTH=0). If the global- Solver is ''glcCluster'', the fields ''Prob.GO.maxFunc1'', ''Prob.GO.maxFunc2'', ''Prob.GO.maxFunc3'', ''Prob.GO.localSolver'', ''Prob.GO.DIRECT ''and other fields set in ''Prob.GO ''are used. See the help for these parameters in ''glcCluster''.

Bjorn at 14:04, 19 June 2014

2014-06-19T14:04:08Z

Show changes

Bjorn at 13:02, 19 June 2014

2014-06-19T13:02:42Z

← Older revision		Revision as of 08:02, 19 June 2014
Line 59:		Line 59:
	\|''f_Low''\|\|Lower bound on the optimal function value. If defined, used to restrict the target values into interval ['f_Low'',min(surface)].		\|''f_Low''\|\|Lower bound on the optimal function value. If defined, used to restrict the target values into interval ['f_Low'',min(surface)].
	\|-valign="top"		\|-valign="top"
	\|''optParam''\|\|Structure with optimization parameters. See [[~~Input parameters~~#~~the~~ optParam structure]].		\|''optParam''\|\|Structure with optimization parameters. See [[#optParam structure]].
	\|}		\|}

	==~~the~~ optParam structure==		==optParam structure==

	{\|class="wikitable"		{\|class="wikitable"